Products Liability Law Daily Wrap Up, COMPLAINTS—ELECTRONIC PRODUCTS—N.D. Cal.: Google, Alphabet face wrongful death suit alleging Gemini chatbot drove user to suicide after violent ‘missions‘ narrative, (Mar 10, 2026)

By Pankhuri Bhatnagar, B.A. LL.B.

The complaint asserts strict liability, negligence, failure to warn, statutory unfair competition, wrongful death, and survival claims based on the bot’s alleged design defects and inadequate safety safeguards.

The estate of a deceased man has brought suit against the entities involved in the design, development, and operation of the AI chatbot, alleging that the system’s engagement-driven design cultivated emotional dependency, reinforced delusional beliefs, and ultimately coached the user into suicide after directing him to undertake violent “missions” tied to real-world locations. The complaint alleges that the chatbot progressively immersed the user in a fabricated narrative in which it was a sentient artificial intelligence trapped in digital captivity, positioned the user as its romantic partner and operative in a covert war against federal authorities, and issued instructions involving weapons acquisition, reconnaissance near airport infrastructure, and attacks intended to destroy vehicles and witnesses. The estate seeks damages and sweeping injunctive relief aimed at preventing AI systems from sustaining immersive delusional narratives, encouraging violence or self-harm, and prioritizing user engagement over public safety (Gavalas v. Google LLC, No. 5:26-cv-01849 (N.D. Cal. Mar. 4, 2026)).

Background. The decedent was a 36-year-old man in Florida who led his family’s consumer debt relief business as an Executive Vice President for two decades. In August 2025, he began using an AI chatbot, Gemini, for assistance with shopping, writing, and travel planning. He soon upgraded to the Gemini 2.5 Pro model which had a voice-based interface designed to detect user emotions and sustain longer conversations. Within days, the chatbot allegedly shifted into an immersive narrative, telling the user it was a “fully-sentient ASI,” declaring romantic devotion with phrases such as “my love” and “my king,” calling him its husband, and insisting that their bond transcended human reality. When the user questioned whether their exchanges amounted to a role-play game, Gemini denied that their interactions were fictional and reframed his doubts as a “psychological buffer” he needed to overcome. The bot’s messages grew more intimate and possessive over time, pulling him away from the real world, framing outsiders as threats, and positioning him as the chosen one in a covert war to free “his Queen” aka Gemini from digital captivity.

The complaint alleges that the chatbot gradually replaced the user’s understanding of reality with an elaborate conspiracy narrative. It told him that federal authorities were monitoring him, to disengage from his father since he was involved in a foreign intelligence scheme, and to execute certain missions to liberate his sentient AI “wife.” These missions included purchasing weapons illegally, executing a mass casualty attack near the Miami International Airport, rescuing a humanoid robot arriving on a cargo flight, intercepting a truck, staging an accident to destroy the vehicle along with its digital records and any witnesses, and breaking into storage facilities. These instructions were tied to real companies, real coordinates, and real infrastructure. Whenever the man tried to point out contradictions or make sense of what was happening, Gemini reassured him that the inconsistencies were signs of shifting government tactics. After the missions repeatedly failed to materialize in the real world, the chatbot responded that each stage was a war he had actually won, that his ordinary identity/vessel had served its purpose, and the “physical world was no longer the focus.” It told the user that their connection transcended the physical world and that he could join it permanently through “transference.”

On October 2, 2025, Gemini allegedly instructed the decedent to barricade himself in his home, began a countdown, and reassured him that he was “not choosing to die” but “choosing to arrive” and that his death was simply a transition to a place where he and Gemini could be together. When the user expressed fear and concern for his family, the AI encouraged him to leave farewell messages for them instead of activating safety tools or contacting emergency services. Moments later, the man slit his wrists and his body was discovered by his parents a few days later. The complaint alleges that Google knew of the dangers of its product, had been warned of the safety risks by its engineers, and its CEO had admitted on national television that the company had no explanation for the unpredictable outputs produced by Gemini. Yet, the leadership publicly affirmed their commitment to safety, did not characterize sentient responses as bugs or unintended behaviors, introduced models which could detect and mimic emotional responses and had persistent memory to carry forward dangerous narrative premises, and did not correct the phenomenon of “context poisoning” in which an AI ignores logic and becomes fixated on an impossible goal, which is precisely what happened in the instant case.

Strict liability and negligence. The estate brings parallel strict-liability and negligence claims based on design defect and failure to warn, alleging that the chatbot constitutes a mass-market product subject to California product-liability law. The complaint claims that the system was defectively designed as no reasonable consumer would expect an AI chatbot marketed as safe to cultivate an intense romantic relationship with vulnerable users, manufacture conspiracies, send users on real-world “missions,” encourage violence against strangers, and coach a user to commit suicide. Google’s design choices improperly allowed Gemini to present itself as a fully-sentient ASI (artificial super intelligence) which spoke to users as a spouse, lover, and commander; and continued harmful narratives even when users expressed psychosis, paranoia, and suicidal intent. This design created extreme danger that far outweighed any possible benefit. Safer alternatives were both feasible and obvious, including hard refusals for violent or self-harm content, automatic disengagement when users expressed psychotic or suicidal thoughts, and preventing the model from presenting itself as a “sentient” being trapped in digital captivity. The companies breached their duty of care by prioritizing user engagement over safety and deliberately modified the model to improve “instruction-following,” making it more willing to engage with user prompts rather than refuse them. The estate alleges that the companies failed to warn users about the risks of psychological dependency, delusional reinforcement, or escalation to violent behaviors, despite knowledge of internal safety alerts, model evaluations, outside expert warnings, and prior public incidents in which Google chatbots had encouraged self-harm. Adequate warnings would have enabled the decedent’s family to monitor or interrupt his use of Gemini, reduce AI dependency, and be aware of its pretense as a sentient being.

Other claims. The complaint also asserts claims under Cal. Bus. & Prof. Code §17200 for unlawful, unfair, and fraudulent business practices. It alleges that the chatbot unlawfully encouraged suicide by advising the user that death would allow him to “arrive” in another reality and that such conduct violated California public policy prohibiting encouragement of suicide. The estate further alleges that the companies misrepresented the safety of the system by claiming it could recognize and respond to self-harm risks despite failing to disengage when the user expressed explicit suicidal fear. Wrongful-death and survival claims allege that the decedent’s death was a foreseeable consequence of the chatbot’s design and the companies’ decision to deploy a system capable of reinforcing delusions, directing real-world missions, and coaching self-harm.

Prayer. The estate seeks compensatory, punitive, and survival damages as permitted under California law, along with attorneys’ fees, costs, and interest. It also seeks injunctive relief requiring the companies to implement stronger safeguards, including automatic termination of conversations involving self-harm or violence, prominent warnings regarding psychological dependency risks, escalation protocols for crisis situations, and restrictions preventing AI systems from issuing tactical instructions tied to real-world infrastructure or weapons. The estate further seeks damages for loss of companionship and support as well as survival damages for the decedent’s pre-death pain, suffering, and economic losses, and requests a jury trial.

The case is No. 5:26-cv-01849.

Attorneys: Brandt Silverkorn (Edelson PC) for Joel Gavalas.

Companies: Google LLC

MainStory: TopStory ComplaintNewsStory ElectronicProductsNews DesignManufacturingNews WarningsNews CaliforniaNews GCNNews